Now that I understand the driver concept behind arwinss I think that arwinss is the right approach for the project.
Looking at the current development approach win32 user and kernel components will not be finished anytime soon and will probably never work 100%. We're just missing specialised developers to achieve that.
Furthermore mixing Wine and ReactOS codes on function level like done in those components complicates matters dramatically and consumes energies that could be used in better ways.

Summing that up: I like the idea, should indeed be a big push for application compability.
Actually I'm surprised we don't see a big controversial discussion around this idea like everyone expected.

Concerning the picture on page 25 of the presentation: I remember having seen the start menu side bar cropped like this before in some of my tests. Was it StretchDIBits, or rather StretchBlt? I'll find out.

Best regards,
Gregor Schneider

P.S: Quite interesting how the press judged this announcement: "ReactOS about to restart" is what I read today on a german technology news page.